Long‐term monitoring is critical for documenting population and community trends and for management, especially in large river‐floodplain ecosystems that provide important services. Levees have reduced active floodplains in most large rivers, but connectivity in some reaches could promote community resilience. Using multivariate tools and regression, we examined spatial and temporal structure in fish community samples from two decades (>1000 samples, >100 sites) in one of the largest relatively intact river‐floodplain ecosystems in North America – the Atchafalaya River (ARB), Louisiana. Assemblages exhibited significant structure temporally and spatially, with most substantial effects of year and hydrologic subdivisions; however, season and water level also influenced community structure. Temporal trends in communities were limited to few areas, while declines in species richness were more widespread; however, rarefied richness trends suggested that declines were slight. Recent and long‐term declines in abundance of economically important species (e.g., Black Crappie, Largemouth Bass, Bigmouth Buffalo) and increases in others (e.g., Smallmouth Buffalo, Blue Catfish, Freshwater Drum) drove differences among time periods. Our results suggest that the hydrologic subdivisions of the ARB may be an appropriate scale at which to manage fish populations, hydrology and water quality. Although we could not account for several important factors affecting fish communities in the ARB (e.g., hurricanes, major floods), and were limited by sampling variability, our findings highlight the utility of long‐term datasets from large river‐floodplain ecosystems for identifying important scales for management, determining species contributions to community change and forming hypotheses about anthropogenic and environmental drivers of variation in fish communities.